Two new schemes for group-wise monitoring of multivariate data and diagnosis of out-of-control individuals

Group-wise monitoring means that when at least one individual within a sample is out-of-control (OC), the corresponding sample is considered OC. Currently, there is much research on group-wise monitoring of univariate data, but relatively limited work exists for multivariate data. Motivated by increasing product complexity and technological advances, this article proposes two new schemes for group-wise monitoring of multivariate data and OC individual diagnosis. Scheme 1 is based on a modified Holm's step-down multiple testing procedure and can directly identify OC individuals; Scheme 2 constructs an adaptive EWMA control chart and uses available OC information with a support vector machine to diagnose OC individuals. Simulation results show that both schemes outperform the existing method in monitoring performance. In terms of diagnosis, Scheme 2 (Scheme 1) performs better in scenarios involving multiple OC individuals (one OC individual). Finally, a gas-meter-motivated simulation study is conducted to demonstrate the advantages of the proposed schemes.